Survey results
Attendee travel survey — what came back, what we extrapolated, and how.
Mode mix (post-extrapolation)
Shares aggregate respondent responses with the statistical projection for non-responders. The projection holds the mode-mix constant within each distance band, not across the whole event — short-haul mix is calculated separately from local mix.
Sample of responses
| Ref | Origin | Mode | Miles | Nights | Band |
|---|---|---|---|---|---|
| R-0042 | SE1, GB | Walk / cycle | 2 | 0 | 0-25 |
| R-0117 | M14, GB | Rail | 198 | 1 | 100-500 |
| R-0128 | EH9, GB | Rail | 412 | 1 | 100-500 |
| R-0149 | CF24, GB | Car (petrol/diesel) | 152 | 1 | 100-500 |
| R-0162 | G2, GB | Short-haul flight | 372 | 1 | 100-500 |
| R-0188 | B15, GB | Rail | 110 | 0 | 100-500 |
| R-0203 | 75008, FR | Short-haul flight | 218 | 1 | 100-500 |
| R-0221 | NW3, GB | Car (electric) | 6 | 0 | 0-25 |
| R-0247 | 1011, NL | Short-haul flight | 230 | 1 | 100-500 |
| R-0259 | BS1, GB | Coach | 118 | 0 | 100-500 |
| R-0274 | 10001, US | Long-haul flight | 3,460 | 2 | 500+ |
| R-0301 | LE1, GB | Car (petrol/diesel) | 105 | 0 | 100-500 |
Distance bands
Bands feed both the survey UI ("how far did you come?") and the extrapolation — we project mode-mix within each band rather than mixing short-haul and long-haul into one bucket, because the factors differ by an order of magnitude.
Non-responder treatment
We do not assume they travelled the same way as respondents on average. We extrapolate the mode-mix within each distance band, then apply that mix to the unknown attendees by best-available distance signal (ticket origin postcode where captured at sale, country-of-purchase otherwise).
The projection is recorded as a separate set of activity records with a source="extrapolated" flag, so the methodology PDF can disclose which portion of the headline is statistically projected and which is direct.
Survey questions and validation mirror the eight v1 transport modes accepted by the real attendee survey API: car, coach, rail, ferry, short-haul flight, long-haul flight, walk/cycle, EV car.
Non-responder imputation — math
148 of 460 attendees (30.7%) did not respond to the post-event survey. Their travel emissions are imputed, not ignored. The imputation method and the additional σ it contributes are documented here so an auditor can verify the assumption.
The respondent mode-mix is projected onto the non-responder count per (mode × distance-band) bucket, multiplied by the bucket's mean miles per attendee. Extrapolated buckets are kept separate from respondent records and carry an additional log-space σ of 0.50 × (1 − response rate), composed in quadrature with the bucket's structural σ. No imputation for accommodation — accommodation is only counted when explicitly reported.
- Imputed kgCO₂e: 2,310
- σ contribution: ± 845 (36.6% σ)
- Cohort mean kg/attendee: 16
- Imputation σ component (95% CI share): 29.1%
- Formula: σimputation = 0.50 × (1 − 0.693) = 0.154 (log space), composed in quadrature
The imputation appears as a discrete line in Cat 6 — Travel rather than being silently rolled into other lines, so an auditor can isolate the imputation contribution independently.